Maximum dissipation principle in numerical simulation of
complex fluids
Yunkyong Hyon
Institute for Mathematics and Its Applications
IMA PI Workshop on Numerical Simulation of Complex Fluids and MHD, Penn. State University
Outline
Complex Fluids
Complex Fluids
Macroscopic fluids with microscopic configuration/structure
Competition between the kinetic energy and the internal
“elastic” energy
Complex Fluids
Macroscopic fluids with microscopic configuration/structure
Competition between the kinetic energy and the internal
“elastic” energy
Complex Fluids
Macroscopic fluids with microscopic configuration/structure
Competition between the kinetic energy and the internal
“elastic” energy
Complex Fluids : Examples
Viscoelastic Fluids
(
Polymeric Materials, Liquid Crystal,...)
Electrorheological (ER) Fluids
(
Ion Dynamics(Electrolyte Solutions) in Biology, Semiconductor,...)
Complex Fluids
The analysis of complex fluids is intrinsically difficult.
Numerical computations are highly demanded to understand the
complex fluids.
The starting point of energetic variational approaches: Energy
dissipation law for the whole coupled system,
d
dt
E
total
=
−△
where
E
totalis the total energy,
△
is the dissipation of the
Complex Fluids
The analysis of complex fluids is intrinsically difficult.
Numerical computations are highly demanded to understand the
complex fluids.
The starting point of energetic variational approaches: Energy
dissipation law for the whole coupled system,
d
dt
E
total
=
−△
where
E
totalis the total energy,
△
is the dissipation of the
Energetic Variational Approaches: Flow Map
b b
~x(X~,t) Ω0
~ X
Ωt
~x
Figure: The flow map from the reference domain, Ω0to the current domain, Ωt.
Basic Mechanics
Flow Map (Trajectory) : ~xt(X~,t) =~u(~x(X~,t),t), ~x(X~,0) =X~.
The deformation tensor (strain) of the flow map :
F(~x(X~,t),t) = ∂~x(X~,t)
∂ ~X satisfying
Energetic Variational Approaches
Least Action Principle (Hamiltons Principle; Principle of Virtual
Work) :
the reversible part of the system, conservative force (
f
c)
δ
A
=
f
c·
δ~
x
where
A
is the action functional defined from the energy.
Maximum Dissipation Principle
1(Onsager’s Principle) :
the
irreversible part of the system, the dissipative force (
f
d)
1
2
δ
△
=
f
d·
δ~
u
.
Energetic Variational Approaches : Simple Fluid
Simple Fluid :
the fluid described by the incompressible Navier-Stokes
equations
ρ
(
~
u
t+
~
u
· ∇
~
u) +
∇
p
=
µ
∆
~
u
,
(momentum conservation)
∇ ·
~
u
= 0
,
(incompressibility)
(2)
ρ
t+
∇ ·
(
ρ~
u
) = 0
.
(mass conservation)
where
ρ
is mass,
~
u
is velocity field,
p
is pressure, and
µ
is the viscosity.
The energy equation for incompressible Navier-Stokes equation:
d
dt
Z
1
2
ρ
|
~
u
|
2
d
~
x
=
−
Z
Energetic Variational Approaches : Simple Fluid
The total energy, the dissipation :
E
total=
Z
1
2
ρ
|
~
u
|
2
d
~
x
,
△
=
Z
Energetic Variational Approaches : Simple Fluid
Define Action Functional,
A
, for
Least Action Principle (LAP)
A
=
Z
T0
Z
Ωt
1
2
ρ
|
~
u
|
2
d
~
xdt
.
(5)
Pull back the current domain, Ω
t, to the reference one, Ω
0, through
the flow map,
~
x
(
X
~
,
t
)
A
(
~
x) =
Z
T0
Z
Ω0
1
2
ρ
0(
X
~
)
det
F
|
~
x
t|
2
det
F d
X dt
~
(6)
where
ρ
0(
X
~
) =
ρ
(
X
~
,
t
)
|
t=0is the initial mass. Then
δA(~x)δ~x
= 0 gives
the Hamiltonian (energy conserved) part under the incompressibility
condition,
∇ ·
~
u
= 0, i.e., det(
F
) = 1.
The resulting equation is the Euler equation : the total energy
conservation
ρ
(
~
u
t+
~
u
· ∇
~
u) =
−∇
p
Energetic Variational Approaches : Simple Fluid
Applying
Maximum Dissipation Principle (MDP)
for the dissipation
in the form of the quadratic of the “rate” functions,
1
2
δ
△
δ~
u
ε=0
=
Z
(
∇
~
u
+
ε
∇
~
v
)
· ∇
~
v d
~
x
ε=0
= 0
,
then we obtain the Stokes equation,
µ
∆
~
u
=
∇
˜
p
,
(8)
∇ ·
~
u
= 0
.
Energetic Variational Approaches : Framework
1
Define an appropriate energy equation for describing the physical
phenomenon.
2
Applying energetic variational approaches, LAP, MDP.
3
Combine the conservative part from LAP and the dissipative part
from MDP through the force balance to satisfy the energy law.(The
system of PDE is uniquely determined.)
4
Numerical Computations, developing numerical methods, existence,
Energetic Variational Approaches : Framework
1
Define an appropriate energy equation for describing the physical
phenomenon.
2
Applying energetic variational approaches, LAP, MDP.
3
Combine the conservative part from LAP and the dissipative part
from MDP through the force balance to satisfy the energy law.(The
system of PDE is uniquely determined.)
4
Numerical Computations, developing numerical methods, existence,
Energetic Variational Approaches : Framework
1
Define an appropriate energy equation for describing the physical
phenomenon.
2
Applying energetic variational approaches, LAP, MDP.
3
Combine the conservative part from LAP and the dissipative part
from MDP through the force balance to satisfy the energy law.(The
system of PDE is uniquely determined.)
4
Numerical Computations, developing numerical methods, existence,
Energetic Variational Approaches : Framework
1
Define an appropriate energy equation for describing the physical
phenomenon.
2
Applying energetic variational approaches, LAP, MDP.
3
Combine the conservative part from LAP and the dissipative part
from MDP through the force balance to satisfy the energy law.(The
system of PDE is uniquely determined.)
4
Numerical Computations, developing numerical methods, existence,
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow
2:
Letφbe a phase functionφ(~x,t) =±1 in the incompressible fluids
Let Γt ={x:φ(~x,t) = 0}be the interface of mixture
In the Eulerian description, the immiscibility of fluids implies the pure transport ofφ.
φt+~u· ∇φ= 0. (9)
2
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow:
Ginzburg-Landau mixing energy, represents a competition between two fluids with their (hydro-) philic and (hydro-) phobic properties:
W(φ,∇φ) = 1 2|∇φ|
2+ 1
4η2(φ 2−1)2.
The equilibrium profile of interface is tanh-like function asη→0.
The total energy is defined by the combination of the kinetic energy and the internal energy as follows:
Etotal =
Z
Ωt
1
2|~u|
2+
λW(φ,∇φ)
d~x. (10)
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow:
The action functionalAfrom the total energy (10):
A= Z T 0 Z Ωt 1 2|~xt|
2
−λW(φ,∇φ)
d~xdt. (11)
The action functionalAin terms of the flow map:
A(~x) =
Z T 0 Z Ω0 1 2|~xt|
2 −λ
1 2|F
−1∇ ~ Xφ|
2+f(φ0)
detF
dX dt.~
(12) wheref(φ0) = 1
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow without Dissipation:
the LAP leads us to the following Hamiltonian system:
~ut+~u· ∇~u+∇p˜ = −λ∇ ·(∇φ⊗ ∇φ) (13)
∇ ·~u = 0 (14)
φt+~u· ∇φ = 0. (15)
The above system (18) – (20) converges to the “sharp interface model” and its energy equation is
d dt
Z 1
2|~u|
2+λ 1
2|∇φ|
2+ 1
4η2(φ 2
−1)2
d~x = 0. (16)
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow:
Consider the following dissipation with the viscosity of fluids and the dissipation within the diffusive interface:
△=
Z
µ|∇~u|2+λγ
∆φ− 1
η2(φ 2
−1)φ
2!
d~x. (17)
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow:
The system with dissipation:
~ut+~u· ∇~u+∇p˜ = µ∆~u−λ∇ ·(∇φ⊗ ∇φ) (18)
∇ ·~u = 0 (19)
φt+~u· ∇φ = γ
∆φ− 1 η2 φ
2 −1
φ
. (20)
The dissipative energy law is
d dt
Z
1 2|~u|
2+λ
1 2|∇φ|
2+ 1
4η2(φ 2
−1)2
d~x
(21) =−
Z
µ|∇~u|2+λγ
∆φ− 1 η2(φ
2 −1)φ
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow with Dissipation:
Reducing the computational cost in numerical experiment through the
Maximum dissipation principle.
Manipulate the dissipation (17) in terms of a rate
△=
Z
µ|∇~u|2+λ
γ|φt+~u· ∇φ| 2d~
x. (22)
Onsarger’s principle (MDP) with incompressibility of flow,∇ ·~u= 0 implies
1 2
δ△ δ~u
ε=0
=− Z
µ∆~u−λ
γ(φt+~u· ∇φ)∇φ
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow:
The resulting system obtained by LAP and MDP with the manipulation on the dissipation is
~ut+~u· ∇~u+∇p = µ∆~u−
λ
γ(φt+~u· ∇φ)∇φ (24)
∇ ·~u = 0 (25)
φt+~u· ∇φ = γ
∆φ− 1 η2(φ
2 −1)φ
. (26)
The dissipative energy law:
d dt
Z
1 2|~u|
2+λ
1 2|∇φ|
2+ 1
4η2(φ 2
−1)2
d~x
(27)
Z
Energetic Variational Approaches : Complex Fluid
Remark:
The dissipative force term in (24),
−λ
γ(φt+~u· ∇φ)∇φ
is exactly same as the conservative force in (18),
∇ ·(∇φ⊗ ∇φ−W(φ,∇φ)I)
Asη →0, this is exactly the surface tension force on the interface3
3
Energetic Variational Approaches : Complex Fluid
Immiscible Two-phase Flow (continue):
Computational Viewpoints:
d dt
Z 1
2|~u|
2+λ 1
2|∇φ|
2+ 1
4η2(φ 2−1)2
d~x
(28) =−
Z
µ|∇~u|2+λ
γ|φt+~u· ∇φ| 2
d~x.
: Lower Order Approximation in Numerical Simulations
d dt
Z 1
2|~u|
2+ λ
1
2|∇φ|
2+ 1
4η2(φ 2
−1)2
d~x
(29) =−
Z
µ|∇~u|2+λγ|∆φ− 1
η2(φ 2
−1)φ|2
Energetic Variational Approaches : Complex Fluid
Another Example:
an incompressible viscoelastic complex fluid
model
4with the elastic energy,
W
(
F
) =
12|
F
|
2The energy law:
d dt
Z
1 2|~u|
2+1
2|F|
2
d~x=− Z
µ|∇~u|2d~x. (30)
The system of equations satisfies the energy law (30):
~ut+~u· ∇~u+∇p = µ∆~u+∇ · WFF−T (31)
∇ ·~u = 0 (32)
Ft+~u· ∇F = ∇~uF (33) : WFF−T is conservative force.
4
Energetic Variational Approaches : Complex Fluid
Another Example (continue):
Add an artificial term ∆
F
into (33)
F
t+
~
u
· ∇
F
=
∇
~
uF
+
ε
2∆
F
.
(34)
The energy law of the system (31), (32), (34):
d dt
Z 1
2|~u|
2+1
2|F|
2
d~x=− Z
µ|∇~u|2+ε2|∇F|2
d~x. (35)
Apply MDP: the resulting energy equation is
d dt
Z 1
2|~u|
2+1
2|F|
2 d~x
(36) =−
Z
Numerical Simulations
Finite Element Space
:
~
u
h∈
V
h= (P
1⊕
bubble)
2(37)
p
h∈
W
h=
P
1(38)
φ
h∈
Q
h=
P
1(39)
Numerical Simulations
Numerical Method : The explicit-implicit scheme
5(˜~uh,tn+1, ~vh) +
3~
uhn−~un−1 h
2 · ∇
~u
n+1 2
h , ~vh
+
1
2
∇ ·3~u
n h−~u
n−1 h 2 ~u n+1 2
h , ~vh
−(p
n+1 2
h ,∇ ·~vh) =−(µ∇~u
n+1 2
h ,∇~vh)− λ γ
˜
φnh,t+1∇ 3φn
h−φ n−1 h
2
, ~vh
(40) −λ γ ~ un+12
h ,∇
3φn h−φn−
1 h
2 ∇
3φn h−φn−
1 h
2
, ~vh
for all~vh∈Vh,
∇ ·~un+12
h ,wh
+εpn+12
h ,wh
= 0 for allwh∈Wh, (41)
( ˜φnh,t+1,qh) +
~ u n+1 2 h ,
3φn h−φ
n−1 h 2 qh (42) =−γ∇φ
n+1 2
h ,∇qh
− γ
η2(fh(φ n
Numerical Simulations
where
fh(φnh, φn +1 h ) =
(
|φnh+1|2−1) + (|φn h|2−1)
2 φ n+1 2 h , ˜ ~
uh,tn+1=~u
n+1 h −~unh
∆t , ~u
n+1 2
h =
~uhn+1+~un h
2 ,φ˜
n+1 h,t =
φnh+1−φnh
∆t , φ
n+1 2
h =
φnh+1+φnh
2 , (·,·) is the inner product operator, and ∆t is the time step for simulations andε
is a small positive constant, 0< ε <<1. The scheme satisfies the energy dissipation law:
Z 1
2|~u
n+1 h |
2+ λ
1
2|∇φ
n+1 h |
2+ 1
4η2|φ n+1 h
2 −1|2
d~x h,t = (43) − Z (
µ|∇~unh+1|2+ε|φn +1 2
h | 2+λ
γ ˜
φnh,t+1+ (~un +1 2
h · ∇)
3φn h−φn−
Numerical Simulations
Simulation 1 - Initial Data
:
~
u
0=
~
0
,
φ
0= tanh
d
1
(x
,
y
)
η
√
2
+ tanh
d
2
(x
,
y
)
η
√
2
−
1
.
0
(44)
with
d
1(x
,
y
) =
q
(x
−
0
.
38)
2+ (y
−
0
.
5)
2−
0
.
11
,
d
2(x
,
y
) =
q
Numerical Simulations
Simulation 1.- Initial Data,
φ
0:
Numerical Simulations
Plot of Interfaces and Flow Fields at Time =0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7
0.8 Plot of Interfaces and Flow Fields at Time =0.1
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7
0.8 Plot of Interfaces and Flow Fields at Time =0.2
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Plot of Interfaces and Flow Fields at Time =0.4
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Plot of Interfaces and Flow Fields at Time =0.7
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7
0.8 Plot of Interfaces and Flow Fields at Time =1
Numerical Simulations
0 0.2 0.4 0.6 0.8 1 1.2 0
0.002 0.004 0.006 0.008 0.01 0.012 0.014
Time : t
Energy
Energy Dissipation : Merging Effect
Total Energy Mixing Energy Kinetic Energy
0 0.2 0.4 0.6 0.8 1 1.2 0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8x 10
−6
Time : t
Energy
Kinetic Energy Evolution : Merging Effect
Numerical Simulations
Simulation 2 - Initial Data
:
~
u
0=
~
0
,
φ
0= tanh
d
1
(x
,
y
)
η
√
2
+ tanh
d
2
(x
,
y
)
η
√
2
−
1
.
0
(45)
with
d
1(x
,
y) =
q
(x
−
0
.
38)
2+ (y
−
0
.
38)
2−
0
.
22
,
d
2(x
,
y) =
q
Numerical Simulations
Simulation 2.- Initial Data,
φ
0:
Numerical Simulations
Plot of Interfaces and Flow Fields at Time =0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.9 Plot of Interfaces and Flow Fields at Time =0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Plot of Interfaces and Flow Fields at Time =0.4
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.9 Plot of Interfaces and Flow Fields at Time =2
Numerical Simulations
0 0.5 1 1.5 2 0
0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Time : t
Energy
Energy Dissipation : Vanishing Effect
Total Energy Mixing Energy Kinetic Energy
0 0.5 1 1.5 2 0
0.5 1 1.5 2 2.5 3x 10
−6
Time : t
Energy
Kinetic Energy Evolution : Vanishing Effect